基于CatBoost算法的硕士研究生就业能力预测模型  

Postgraduate employment forecast based on CatBoost algorithm

作　　者：巩红[1] 陈阳 周晨晖 李昊楠 喻小康 GONG Hong;CHEN Yang;ZHOU Chenhui;LI Haonan;YU Xiaokang(Graduate School,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;School of Economics and Management,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;School of Marxism,Xi'an University of Posts and Telecommunications,Xi'an 710121,China)

机构地区：[1]西安邮电大学研究生院,陕西西安710121 [2]西安邮电大学经济与管理学院,陕西西安710121 [3]西安邮电大学马克思主义学院,陕西西安710121

出　　处：《西安邮电大学学报》2021年第6期89-96,共8页Journal of Xi’an University of Posts and Telecommunications

基　　金：中国学位与研究生教育学会项目(2020MSA97)。

摘　　要：为了预测硕士研究生的就业能力,构建一种基于CatBoost算法的硕士研究生就业能力模型。首先,选取关于硕士研究生在校期间的图书阅读量、专利、技能证书等31项影响因素数据,采用SMOTE过采样方法处理数据集的不平衡问题。其次,通过机器学习方法挖掘学生个人培养数据与就业之间的关系,利用CatBoost算法构建硕士研究生就业预测模型,并使用10倍交叉法降低结果的偶然性。最后,将基于CatBoost算法的预测模型与随机森林、决策树及支持向量机等相关算法进行比较。研究结果表明,该预测模型在召回率、精确率、误判率、F_(1)值及对硕士研究生的就业能力预测效果方面均优于其他算法。In order to predict the employability of postgraduate students,an employability model based on CatBoost algorithm is constructed.Firstly,the data of 31 influencing factors such as book reading,patents and skill certificates of postgraduates are selected,and the SMOTE oversampling method is adopted to deal with the imbalance of the data set.Secondly,the relationship between students'personal training data and employment is mined through machine learning method,the employment prediction model of postgraduates is constructed by CatBoost algorithm,and the 10x crossover method is used to reduce the contingency of the results.Finally,the prediction model based on CatBoost algorithm is compared with relevant algorithms such as random forest,decision tree and support vector machine.The results show that the prediction model is superior to other algorithms in recall rate,accuracy rate,misjudgment rate,F1 value and the prediction effect on the employability of postgraduates.

关 键 词：就业预测 CatBoost算法 SMOTE过采样 10倍交叉验证 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]